Introduction
The Catalogue of Endangered Languages (ELCat) is one of several similar responses to
a perceived need for better data on language vitality. My remarks here are framed as
a direct reply to Lee & Van Way's article, but really address
larger issues in the ongoing debate about a perceived need to classify, inventory,
and enumerate endangered languages. Lee & Van Way focus on one aspect of
ELCat, the Language Endangerment Index (LEI), discussing a number of shortcomings in
other current models. As an instrument for determining the level of language
endangerment, the LEI is presented as a preferable alternative to other metrics,
including Fishman's (Reference Fishman1991) Graded
Intergenerational Disruption Scale (GIDS), or EGIDS, the Expanded scale, based on
the work of Lewis & Simons (Reference Lewis and Simons2010), or UNESCO's (Reference Brenzinger, Dwyer, de Graaf, Grinevald, Krauss, Miyaoka, Ostler, Sakiyama, Villalón, Yamamoto and Zapeda2003) expert scale. Lee & Van Way's discussion
presupposes that such metrics are needed, and that it is beneficial to have a method
for measuring vitality. Specifically, they argue that ‘for those
concerned with preserving the world's fragile linguistic diversity, it is
desirable to be able to quantify language vitality’. This is the
underlying assumption of not only ELCat and LEI, but of other language catalogues,
such as the Ethnologue (Lewis, Simons, & Fennig Reference Lewis, Simons and Fennig2015), UNESCO's Atlas (Moseley
Reference Moseley2010), and other vitality metrics, as
discussed in Lee & Van Way.
My remarks here are aimed at examining the benefits and hazards of any kind of
numerical scale, the advantages and potential pitfalls of using a catalogue like
ELCat or the Ethnologue, and to consider what we, as linguists,
could be doing differently and better. All the measurements of vitality are
quantitative and reduce complex social dynamics to a single number or set of
numbers. I focus on measurements of vitality rather than endangerment as is the
standard in current discourse of the indigenous communities I work with, primarily
in the Arctic but elsewhere. This represents a significant shift in the
representation of attitudes toward the rhetoric of indigenous languages to one away
from dire predictions about endangerment to action-oriented attitudes about vitality
and sustainability.Footnote 1 This represents a significant and important shift in attitudes that can be
promoted or inhibited by the assessments of linguists, policy makers, and educators.
ELCat represents a tremendous and ambitious effort to provide better data. It is an
ongoing effort that involves a large number of linguists and indigenous language
organizations, and represents a serious commitment to responding to input from both
the linguistic and speaker communities to make better data more broadly available to
users. The LEI is an attempt to provide a more accurate measure of vitality. This
effort is to be applauded, but at the same time it is important to examine both the
LEI and ELCat as a whole within the larger context of measurements and catalogues of
languages.
Lee & Van Way's discussion of LEI is framed by what they see as
a fundamental question in the study of language vitality: ‘How should the
level of endangerment of any language be assessed?’ The decision of the
linguists behind LEI is that four criteria are central (and sufficient) to
determining language vitality: intergenerational transmission, absolute numbers of
speakers, increasing or decreasing numbers of speakers, and domains of use. The LEI
metric differs from UNESCO's measure in weighting different factors
differently, which is an important advantage of LEI. But despite the
article's title, ELCat does not assess language vitality;
rather, it provides a compilation of several factors that have been deemed by the
linguistic community at large to be key in providing information about vitality.
This is the essence of LEI, and thus the claim that LEI measures language
vitality is not strictly accurate. The LEI ‘measurement’ is an
aggregate of assessment data from other sources. LEI itself is the result of
calculated decisions about what constitutes a sign of vitality, and at best provides
a snapshot view of the vitality of (what is identified as) a single language at a
moment in time. In assessing the overall contributions of ELCat and LEI, it is
important to have a clear understanding of just what they do represent and what they
do not.
Enumeration and commodification
Enumeration is the underlying organizing principle of ELCat (as it is for other
language vitality catalogues), and LEI provides a vitality measurement that is
reduced to a single figure, and rated numerically for reliability. Five advantages
to such quantitative measurements are given by Lee & Van Way: (i)
presenting a bird's eye view of vitality, (ii) raising awareness of
endangerment in communities, (iii) providing an assessment tool for the urgency of
action, (iv) enabling cross-linguistic comparison of language vitality, and (v)
facilitating comparisons between linguistic and biodiversity.
Discussions of the pros and cons of assessing endangered language vitality in this
way generally center on three problematic aspects of the very enterprise of
enumeration: (i) the commodification of language, (ii) the problems of classifying
and counting speakers, and (iii) the very act of reducing complex social situations
to a number. None of these points is novel, and I reference deep scholarly debates
on all of them (Hill Reference Hill2002; Dobrin, Austin,
& Nathan Reference Dobrin, Austin, Nathan and Austin2009; Moore,
Pietikäinen, & Blommaert Reference Moore, Pietikäinen; and Blommaert2010). Nonetheless, in examining any resource for languages, it is
useful to revisit them in considering the advantages or disadvantages to LEI and
ELCat, as they lie at the core of the enterprise of classifying and cataloguing
languages and their vitality. Enumeration presupposes that languages and speakers
are bounded, stand-alone entities without (significant) differences due to
variation, fluency levels, and multilingualism. Critically, endangered language
catalogues are the backbone of a larger linguistic ideology that commodifies,
tallies, and audits languages as uniquely identifiable, ‘each with a
formally stable reality and a naturalized inherence in an ethnic group that is
typically described as tied, through culture, to a unique ecological
niche’ (Dobrin et al. Reference Dobrin, Austin, Nathan and Austin2009:38).
This view transforms language into a marketable commodity (Heller Reference Heller2003), and has very real repercussions for how
such major issues as to how authenticity and legitimacy are construed (Bourdieu
Reference Bourdieu1982) and for on-the-ground realities,
such as how resources are allocated and what kinds of policies are implemented by
language planners and governments.
Enumeration in ELCat
A brief look at random entries in ELCat provides a representative view of what I
mean by saying that enumeration is its organizing principle. Click on the
Languages tab and a page opens listing ‘176 Countries/territories
with ELCat listings’.Footnote 2 Click on the first entry and a page appears with a header:
‘23 languages in Afghanistan’.Footnote 3 Here again, click on the first entry (Ashkun) and a page appears stating
that the language is threatened with twenty percent certainty based on six
sources, and the reference for each source is helpfully given, with links to
digital sources where available. The Compare Sources tab presents an overview of
each of these sources in terms of a variety of categories, including number of
speakers and year of the source. Here there are differences: Moseley (Reference Moseley2010) lists 2,000 speakers, and the
sixteenth edition of the Ethnologue (Lewis Reference Lewis2009) is cited as listing 1,200 speakers
from a source dated 2000. ELCat's table does not include the primary
source of the data, but a look at the more recent edition of the
Ethnologue (Lewis et al. Reference Lewis, Simons and Fennig2015) cites 4,000 monolingual speakers out of a population of 40,000
as of 2011, with no source listed for either edition. While it does seem that
ELCat's conclusion, that there are less than 7,000 speakers of Ashkun
worldwide, is reasonable, it is still hard to interpret what the numbers mean.
Moreover, the prime sources for data are two other catalogues; the benefit here
is not new data (Lewis and colleagues Reference Lewis, Simons and Fennig2015 appear to provide that, although their source is not given), but
rather the comparison of sources.
What do these numbers tell us? Do they indicate comprehensiveness of the
catalogue? Or do they amount to an alarmist statement about the number of
countries that have endangered languages? To contextualize this, there are 196
countries in the world, if Taiwan is included,Footnote 4 so 90% of the world's countries have endangered
languages spoken. How does an ordinary user interpret the data?
The hazards of aggregate data
One central claim of Lee & Van Way is that quantitative measures do
provide a ‘bird's eye view’ of something, but it is
not always clear what that something is, and that view can at times be very
misleading. Aggregate data are determined by generalizing over local-level
differences. By their very nature and intent, they are reductionist. That is, the
benefit of a bird's eye view comes at the cost of detailed analysis, and
differences between communities are hidden in the aggregate. Arguably, aggregate
data have the potential benefit of capturing large trends, but they have the great
disadvantage of generalizing and thus hiding differences. To the extent that
linguists and speaker communities are interested in counteracting endangerment
trends, surely information about language vitality should also be of value.
Aggregate measures assume a homogeneity across speech communities that does not
exist. They represent all groups who speak a given language X as if there are no
differences (social or regional) across different groups of a single language. A
concrete example is the case of Evenki vitality in Siberia and the Russian Far
North. As a general trend, it is safe to say that language vitality is stronger in
the eastern part of Siberia, in the Republic of Sakha/Yakutia, than it is in
Krasnoyarsk Krai, a more western part closer to Moscow (Grenoble Reference Grenoble, Ramallo and Farfán2010). To be sure, the overall trend is one of
sharp decline and Evenki is unambiguously endangered, and arguably more endangered
in the western regions than in the east. But even here, data collected in 2008 in
Krasnoyarsk Krai show that in some villages the language is completely lost, while
in others children continue to speak it. Here it may be more useful to study the
differences between vitality to understand what makes some speakers more resistant
to shift than to focus on different levels of endangerment. Yet existing metrics
rest on quantitative data, not qualitative, and more than numbers is required to
understand complex social systems. One of the obvious gaps is the role of language
attitudes, both toward the minority or local language(s) and toward the majority
language(s).
In fact, vitality often varies across smaller groupings of a speaker community: the
vitality of a given language can vary in rural versus urban settings or from village
to village. For example, in Daghestan where a strong sense of ethnolinguistic
identity and pride has long ensured language vitality, urbanization of some groups
has meant that indigenous languages are robustly spoken only in relatively isolated
mountain communities, and speakers in the cities are rapidly shifting to Russian
(Samojlova Reference Samojlova2009; Magomedov Reference Magomedov2014). A study of language vitality conducted
in the 1990s in the Russian North (Vakhtin Reference Vakhtin1992, Reference Vakhtin2001) shows marked variation
in vitality levels from village to village, indicating that vitality, and thus shift
and attrition, need to be investigated at the microlevel, within microecologies.
There are individual differences across neighboring villages in terms of fluency
levels of the target language, the numbers of languages in contact and levels of
multilingualism, and sociolinguistic variables affecting language vitality.
What this means is that in some places there are very fluent speakers (of the same
language and often of the same dialect) and not in others. Missing these differences
can mean missing research opportunities, for studying the languages involved and for
studying the social factors at work in shift or vitality. For this reason, in the
last decade or so, indigenous peoples in the Arctic have advocated that vitality
surveys be conducted in more localized ways, with individual communities doing their
own assessment, and/or individual researchers gathering data on the communities in
which they work (Grenoble & Olsen Reference Grenoble and Olsen (Puju)2014). Without fine-grained knowledge about language communities, it is
really impossible to say what we are comparing. Anecdotally, Arctic indigenous
peoples report similar variation in vitality today, with vast differences even
between neighboring villages. If a snapshot is based on first-hand knowledge of one
village, with vitality for other communities calculated on the assumption that it is
roughly similar, the reality could be very different from the figures put forth.
One concern with all of these numbers is the question of who is included in speaker
counts and how proficiency is determined. Speaker counts rarely reflect
multilingualism and varying proficiency levels in different languages, a range in
proficiency that may correlate with educational levels in any given language or
languages and have nothing to do with language attrition. People living in
multilingual societies may be very fluent in one language in certain domains, and
not proficient in the same language in other domains. Lee & Van Way point
out that the term speakers, when used in the LEI, includes native
speakers as well as semi-speakers and heritage speakers. All of these labels are
problematic, and a linguist's perceptions of what constitutes one or the
other category may not coincide with a community's understanding.
Moreover, these are relative notions, and may shift over time as vitality changes
(Evans Reference Evans, Newman and Ratliff2001:260). There are a number of
different criteria for determining who counts as a speaker, and different ideas
about who has the authority to decide (Moore et al. Reference Moore, Pietikäinen; and Blommaert2010:11–14).
Unless the data come from actual proficiency assessments, numbers may be wildly
inaccurate from a linguist's viewpoint. Even then, the linguistic view
and the community view may clash. Consider the examples of Inuit and Saami, for
example. These are arguably either two languages (each with multiple dialects) or
two language families (which also have multiple dialects). The Inuit Circumpolar
Council (ICC), an international organization representing Inuit living in Canada,
Greenland, the Russian Federation and the US, recognizes a single Inuit
language (Inuit Circumpolar Council 2009);
the Saami Council argues that the Saami people are ‘one people,
united in our own language, culture and history’ (Saami
Political Program 1986). These ideologies
blur the lines between language distinctions in the interests of ethnolinguistic,
cultural, and political unity. To ICC and the Saami Council, numbers matter: the
higher numbers give them a kind of political legitimacy and advocacy in
international institutions like the Arctic Council. ICC puts the total count of
Inuit worldwide at 160,000,Footnote 5 which they see as too large a number to be ignored. The Saami Council makes
similar arguments.
The opposite side of the same coin is the issue of multilingualism. Taking Saami
varieties in Finland as an example, Moore and colleagues (Reference Moore, Pietikäinen; and Blommaert2010) provide a different view of this situation: there are
no monolingual speakers of three different Saami varieties spoken there (Northern
Saami, Skolt Saami, and Inari Saami). Rather, there are speakers for whom one or the
other variety exists as part of a multilingual repertoire. In highly multilingual
societies, language mixing may arguably be the norm, but this means that imposing
rigid definitional boundaries is misleading at best.
Finally, I would add that snapshot, aggregate views of language vitality should not
be used directly for cross-comparison of language vitalities, as Lee &
Van Way suggest, because speaker counts are based on radically different datasets,
gathered at different points of time and by different means. At best, they provide
an indication of where research in such cross comparisons might be fruitfully
conducted. It is misleading to compare vitality for one language from government
census data versus the vitality of a language as determined by a
linguist's first-hand field report, or census data from differing years
as conducted by different governments.
The most widely used data for any catalogue are census data, for very practical
reasons: national censuses tend to provide the most comprehensive information
available. Census data generally relies on self-reporting, and it is often not clear
how respondents to census questionnaires understand the questions and what they are
actually reporting. The quality of census data varies greatly from country to
country. One of the strengths of ELCat is the attempt to mitigate census data by
adding other data points. But still, we are often in a position of relying solely on
official data, as in the case of Sentinelese (see Lee & Van Way), where
the only available data are outdated census figures.
Conclusion
Catalogues are useful resources for collecting and organizing data about languages
and linguistic practices. We live ‘in an age of statistics’
(Urla Reference Urla1993) and are deeply embedded in
cultural practices that make use of counting measures for allocating resources.
There is considerable scholarly debate about the validity of enumeration of
languages (see Hill Reference Hill2002; Heller Reference Heller2003; Moore et al. Reference Moore, Pietikäinen; and Blommaert2010 to name just a few). Yet in today's world,
the use of numbers is inevitable (Austin & Sallabank Reference Austin, Sallabank, Austin and Sallabank2011), and it is certainly better to have a
good catalogue than a bad one. The recent announcement of the decision to make
Ethnologue subscription-based (Lewis Reference Lewis2015) increases the importance of having a high-quality,
open-access catalogue like ELCat. Even the best catalogue should be used with great
caution, however. Numbers present certain dangers and tell, at best, only part of a
very complicated story. Lee & Van Way point to the use of these numbers
by funding agencies to allocate resources for language documentation.
Finally, we should be mindful of the full range of possible ramifications of the uses
of such measures. Anecdotally, grant applications for language documentation work
have been denied because a referee has consulted a catalogue that lists a particular
language as vital when the researcher claims it is not. It is hard to know how to
evaluate such anecdotes, but certainly the potential hazards of inaccurate data can
have very real consequences for how funding is allocated. And for speaker
communities the repercussions can go beyond funding to very real consequences of
language policy, attitudes, and political decisions about identity. ‘As
part of a modern regime of truth that equates knowledge with measurement, statistics
occupy a place of authority in contemporary modes of social description; they are
technologies of truth production. Consequently, at stake in minority concerns with
statistics are not only competing claims to resources but also competing claims to
truth’ (Urla Reference Urla1993:891).
You have
Access
Introduction
The Catalogue of Endangered Languages (ELCat) is one of several similar responses to a perceived need for better data on language vitality. My remarks here are framed as a direct reply to Lee & Van Way's article, but really address larger issues in the ongoing debate about a perceived need to classify, inventory, and enumerate endangered languages. Lee & Van Way focus on one aspect of ELCat, the Language Endangerment Index (LEI), discussing a number of shortcomings in other current models. As an instrument for determining the level of language endangerment, the LEI is presented as a preferable alternative to other metrics, including Fishman's (Reference Fishman1991) Graded Intergenerational Disruption Scale (GIDS), or EGIDS, the Expanded scale, based on the work of Lewis & Simons (Reference Lewis and Simons2010), or UNESCO's (Reference Brenzinger, Dwyer, de Graaf, Grinevald, Krauss, Miyaoka, Ostler, Sakiyama, Villalón, Yamamoto and Zapeda2003) expert scale. Lee & Van Way's discussion presupposes that such metrics are needed, and that it is beneficial to have a method for measuring vitality. Specifically, they argue that ‘for those concerned with preserving the world's fragile linguistic diversity, it is desirable to be able to quantify language vitality’. This is the underlying assumption of not only ELCat and LEI, but of other language catalogues, such as the Ethnologue (Lewis, Simons, & Fennig Reference Lewis, Simons and Fennig2015), UNESCO's Atlas (Moseley Reference Moseley2010), and other vitality metrics, as discussed in Lee & Van Way.
My remarks here are aimed at examining the benefits and hazards of any kind of numerical scale, the advantages and potential pitfalls of using a catalogue like ELCat or the Ethnologue, and to consider what we, as linguists, could be doing differently and better. All the measurements of vitality are quantitative and reduce complex social dynamics to a single number or set of numbers. I focus on measurements of vitality rather than endangerment as is the standard in current discourse of the indigenous communities I work with, primarily in the Arctic but elsewhere. This represents a significant shift in the representation of attitudes toward the rhetoric of indigenous languages to one away from dire predictions about endangerment to action-oriented attitudes about vitality and sustainability.Footnote 1 This represents a significant and important shift in attitudes that can be promoted or inhibited by the assessments of linguists, policy makers, and educators.
ELCat represents a tremendous and ambitious effort to provide better data. It is an ongoing effort that involves a large number of linguists and indigenous language organizations, and represents a serious commitment to responding to input from both the linguistic and speaker communities to make better data more broadly available to users. The LEI is an attempt to provide a more accurate measure of vitality. This effort is to be applauded, but at the same time it is important to examine both the LEI and ELCat as a whole within the larger context of measurements and catalogues of languages.
Lee & Van Way's discussion of LEI is framed by what they see as a fundamental question in the study of language vitality: ‘How should the level of endangerment of any language be assessed?’ The decision of the linguists behind LEI is that four criteria are central (and sufficient) to determining language vitality: intergenerational transmission, absolute numbers of speakers, increasing or decreasing numbers of speakers, and domains of use. The LEI metric differs from UNESCO's measure in weighting different factors differently, which is an important advantage of LEI. But despite the article's title, ELCat does not assess language vitality; rather, it provides a compilation of several factors that have been deemed by the linguistic community at large to be key in providing information about vitality. This is the essence of LEI, and thus the claim that LEI measures language vitality is not strictly accurate. The LEI ‘measurement’ is an aggregate of assessment data from other sources. LEI itself is the result of calculated decisions about what constitutes a sign of vitality, and at best provides a snapshot view of the vitality of (what is identified as) a single language at a moment in time. In assessing the overall contributions of ELCat and LEI, it is important to have a clear understanding of just what they do represent and what they do not.
Enumeration and commodification
Enumeration is the underlying organizing principle of ELCat (as it is for other language vitality catalogues), and LEI provides a vitality measurement that is reduced to a single figure, and rated numerically for reliability. Five advantages to such quantitative measurements are given by Lee & Van Way: (i) presenting a bird's eye view of vitality, (ii) raising awareness of endangerment in communities, (iii) providing an assessment tool for the urgency of action, (iv) enabling cross-linguistic comparison of language vitality, and (v) facilitating comparisons between linguistic and biodiversity.
Discussions of the pros and cons of assessing endangered language vitality in this way generally center on three problematic aspects of the very enterprise of enumeration: (i) the commodification of language, (ii) the problems of classifying and counting speakers, and (iii) the very act of reducing complex social situations to a number. None of these points is novel, and I reference deep scholarly debates on all of them (Hill Reference Hill2002; Dobrin, Austin, & Nathan Reference Dobrin, Austin, Nathan and Austin2009; Moore, Pietikäinen, & Blommaert Reference Moore, Pietikäinen; and Blommaert2010). Nonetheless, in examining any resource for languages, it is useful to revisit them in considering the advantages or disadvantages to LEI and ELCat, as they lie at the core of the enterprise of classifying and cataloguing languages and their vitality. Enumeration presupposes that languages and speakers are bounded, stand-alone entities without (significant) differences due to variation, fluency levels, and multilingualism. Critically, endangered language catalogues are the backbone of a larger linguistic ideology that commodifies, tallies, and audits languages as uniquely identifiable, ‘each with a formally stable reality and a naturalized inherence in an ethnic group that is typically described as tied, through culture, to a unique ecological niche’ (Dobrin et al. Reference Dobrin, Austin, Nathan and Austin2009:38). This view transforms language into a marketable commodity (Heller Reference Heller2003), and has very real repercussions for how such major issues as to how authenticity and legitimacy are construed (Bourdieu Reference Bourdieu1982) and for on-the-ground realities, such as how resources are allocated and what kinds of policies are implemented by language planners and governments.
Enumeration in ELCat
A brief look at random entries in ELCat provides a representative view of what I mean by saying that enumeration is its organizing principle. Click on the Languages tab and a page opens listing ‘176 Countries/territories with ELCat listings’.Footnote 2 Click on the first entry and a page appears with a header: ‘23 languages in Afghanistan’.Footnote 3 Here again, click on the first entry (Ashkun) and a page appears stating that the language is threatened with twenty percent certainty based on six sources, and the reference for each source is helpfully given, with links to digital sources where available. The Compare Sources tab presents an overview of each of these sources in terms of a variety of categories, including number of speakers and year of the source. Here there are differences: Moseley (Reference Moseley2010) lists 2,000 speakers, and the sixteenth edition of the Ethnologue (Lewis Reference Lewis2009) is cited as listing 1,200 speakers from a source dated 2000. ELCat's table does not include the primary source of the data, but a look at the more recent edition of the Ethnologue (Lewis et al. Reference Lewis, Simons and Fennig2015) cites 4,000 monolingual speakers out of a population of 40,000 as of 2011, with no source listed for either edition. While it does seem that ELCat's conclusion, that there are less than 7,000 speakers of Ashkun worldwide, is reasonable, it is still hard to interpret what the numbers mean. Moreover, the prime sources for data are two other catalogues; the benefit here is not new data (Lewis and colleagues Reference Lewis, Simons and Fennig2015 appear to provide that, although their source is not given), but rather the comparison of sources.
What do these numbers tell us? Do they indicate comprehensiveness of the catalogue? Or do they amount to an alarmist statement about the number of countries that have endangered languages? To contextualize this, there are 196 countries in the world, if Taiwan is included,Footnote 4 so 90% of the world's countries have endangered languages spoken. How does an ordinary user interpret the data?
The hazards of aggregate data
One central claim of Lee & Van Way is that quantitative measures do provide a ‘bird's eye view’ of something, but it is not always clear what that something is, and that view can at times be very misleading. Aggregate data are determined by generalizing over local-level differences. By their very nature and intent, they are reductionist. That is, the benefit of a bird's eye view comes at the cost of detailed analysis, and differences between communities are hidden in the aggregate. Arguably, aggregate data have the potential benefit of capturing large trends, but they have the great disadvantage of generalizing and thus hiding differences. To the extent that linguists and speaker communities are interested in counteracting endangerment trends, surely information about language vitality should also be of value.
Aggregate measures assume a homogeneity across speech communities that does not exist. They represent all groups who speak a given language X as if there are no differences (social or regional) across different groups of a single language. A concrete example is the case of Evenki vitality in Siberia and the Russian Far North. As a general trend, it is safe to say that language vitality is stronger in the eastern part of Siberia, in the Republic of Sakha/Yakutia, than it is in Krasnoyarsk Krai, a more western part closer to Moscow (Grenoble Reference Grenoble, Ramallo and Farfán2010). To be sure, the overall trend is one of sharp decline and Evenki is unambiguously endangered, and arguably more endangered in the western regions than in the east. But even here, data collected in 2008 in Krasnoyarsk Krai show that in some villages the language is completely lost, while in others children continue to speak it. Here it may be more useful to study the differences between vitality to understand what makes some speakers more resistant to shift than to focus on different levels of endangerment. Yet existing metrics rest on quantitative data, not qualitative, and more than numbers is required to understand complex social systems. One of the obvious gaps is the role of language attitudes, both toward the minority or local language(s) and toward the majority language(s).
In fact, vitality often varies across smaller groupings of a speaker community: the vitality of a given language can vary in rural versus urban settings or from village to village. For example, in Daghestan where a strong sense of ethnolinguistic identity and pride has long ensured language vitality, urbanization of some groups has meant that indigenous languages are robustly spoken only in relatively isolated mountain communities, and speakers in the cities are rapidly shifting to Russian (Samojlova Reference Samojlova2009; Magomedov Reference Magomedov2014). A study of language vitality conducted in the 1990s in the Russian North (Vakhtin Reference Vakhtin1992, Reference Vakhtin2001) shows marked variation in vitality levels from village to village, indicating that vitality, and thus shift and attrition, need to be investigated at the microlevel, within microecologies. There are individual differences across neighboring villages in terms of fluency levels of the target language, the numbers of languages in contact and levels of multilingualism, and sociolinguistic variables affecting language vitality.
What this means is that in some places there are very fluent speakers (of the same language and often of the same dialect) and not in others. Missing these differences can mean missing research opportunities, for studying the languages involved and for studying the social factors at work in shift or vitality. For this reason, in the last decade or so, indigenous peoples in the Arctic have advocated that vitality surveys be conducted in more localized ways, with individual communities doing their own assessment, and/or individual researchers gathering data on the communities in which they work (Grenoble & Olsen Reference Grenoble and Olsen (Puju)2014). Without fine-grained knowledge about language communities, it is really impossible to say what we are comparing. Anecdotally, Arctic indigenous peoples report similar variation in vitality today, with vast differences even between neighboring villages. If a snapshot is based on first-hand knowledge of one village, with vitality for other communities calculated on the assumption that it is roughly similar, the reality could be very different from the figures put forth.
One concern with all of these numbers is the question of who is included in speaker counts and how proficiency is determined. Speaker counts rarely reflect multilingualism and varying proficiency levels in different languages, a range in proficiency that may correlate with educational levels in any given language or languages and have nothing to do with language attrition. People living in multilingual societies may be very fluent in one language in certain domains, and not proficient in the same language in other domains. Lee & Van Way point out that the term speakers, when used in the LEI, includes native speakers as well as semi-speakers and heritage speakers. All of these labels are problematic, and a linguist's perceptions of what constitutes one or the other category may not coincide with a community's understanding. Moreover, these are relative notions, and may shift over time as vitality changes (Evans Reference Evans, Newman and Ratliff2001:260). There are a number of different criteria for determining who counts as a speaker, and different ideas about who has the authority to decide (Moore et al. Reference Moore, Pietikäinen; and Blommaert2010:11–14).
Unless the data come from actual proficiency assessments, numbers may be wildly inaccurate from a linguist's viewpoint. Even then, the linguistic view and the community view may clash. Consider the examples of Inuit and Saami, for example. These are arguably either two languages (each with multiple dialects) or two language families (which also have multiple dialects). The Inuit Circumpolar Council (ICC), an international organization representing Inuit living in Canada, Greenland, the Russian Federation and the US, recognizes a single Inuit language (Inuit Circumpolar Council 2009); the Saami Council argues that the Saami people are ‘one people, united in our own language, culture and history’ (Saami Political Program 1986). These ideologies blur the lines between language distinctions in the interests of ethnolinguistic, cultural, and political unity. To ICC and the Saami Council, numbers matter: the higher numbers give them a kind of political legitimacy and advocacy in international institutions like the Arctic Council. ICC puts the total count of Inuit worldwide at 160,000,Footnote 5 which they see as too large a number to be ignored. The Saami Council makes similar arguments.
The opposite side of the same coin is the issue of multilingualism. Taking Saami varieties in Finland as an example, Moore and colleagues (Reference Moore, Pietikäinen; and Blommaert2010) provide a different view of this situation: there are no monolingual speakers of three different Saami varieties spoken there (Northern Saami, Skolt Saami, and Inari Saami). Rather, there are speakers for whom one or the other variety exists as part of a multilingual repertoire. In highly multilingual societies, language mixing may arguably be the norm, but this means that imposing rigid definitional boundaries is misleading at best.
Finally, I would add that snapshot, aggregate views of language vitality should not be used directly for cross-comparison of language vitalities, as Lee & Van Way suggest, because speaker counts are based on radically different datasets, gathered at different points of time and by different means. At best, they provide an indication of where research in such cross comparisons might be fruitfully conducted. It is misleading to compare vitality for one language from government census data versus the vitality of a language as determined by a linguist's first-hand field report, or census data from differing years as conducted by different governments.
The most widely used data for any catalogue are census data, for very practical reasons: national censuses tend to provide the most comprehensive information available. Census data generally relies on self-reporting, and it is often not clear how respondents to census questionnaires understand the questions and what they are actually reporting. The quality of census data varies greatly from country to country. One of the strengths of ELCat is the attempt to mitigate census data by adding other data points. But still, we are often in a position of relying solely on official data, as in the case of Sentinelese (see Lee & Van Way), where the only available data are outdated census figures.
Conclusion
Catalogues are useful resources for collecting and organizing data about languages and linguistic practices. We live ‘in an age of statistics’ (Urla Reference Urla1993) and are deeply embedded in cultural practices that make use of counting measures for allocating resources. There is considerable scholarly debate about the validity of enumeration of languages (see Hill Reference Hill2002; Heller Reference Heller2003; Moore et al. Reference Moore, Pietikäinen; and Blommaert2010 to name just a few). Yet in today's world, the use of numbers is inevitable (Austin & Sallabank Reference Austin, Sallabank, Austin and Sallabank2011), and it is certainly better to have a good catalogue than a bad one. The recent announcement of the decision to make Ethnologue subscription-based (Lewis Reference Lewis2015) increases the importance of having a high-quality, open-access catalogue like ELCat. Even the best catalogue should be used with great caution, however. Numbers present certain dangers and tell, at best, only part of a very complicated story. Lee & Van Way point to the use of these numbers by funding agencies to allocate resources for language documentation.
Finally, we should be mindful of the full range of possible ramifications of the uses of such measures. Anecdotally, grant applications for language documentation work have been denied because a referee has consulted a catalogue that lists a particular language as vital when the researcher claims it is not. It is hard to know how to evaluate such anecdotes, but certainly the potential hazards of inaccurate data can have very real consequences for how funding is allocated. And for speaker communities the repercussions can go beyond funding to very real consequences of language policy, attitudes, and political decisions about identity. ‘As part of a modern regime of truth that equates knowledge with measurement, statistics occupy a place of authority in contemporary modes of social description; they are technologies of truth production. Consequently, at stake in minority concerns with statistics are not only competing claims to resources but also competing claims to truth’ (Urla Reference Urla1993:891).